Mobile MCP

npx -y @mobilenext/mobile-mcp@latest

{
  "mcpServers": {
    "mobile-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@mobilenext/mobile-mcp@latest"
      ]
    }
  }
}

Most mobile agents work from screenshots, which means a vision model, image tokens on every step, and coordinates that break on the next layout change. This server reads the native accessibility tree instead and only falls back to screenshots and coordinate taps when an element is not exposed, which makes runs cheaper and the output structured rather than guessed. One tool surface covers both platforms — iOS simulators, Android emulators and real devices over adb or go-ios — so no XCUITest or Espresso knowledge is needed. Beyond taps and swipes it installs, launches and terminates apps, records the screen, opens deep links, sets orientation and reads crash logs.